The potential reduction in surface runoff loss of band-applied herbicides was studied in a rainfall simulation. Concentration and loss of atrazine, metribuzin, and metolachlor applied in a band 25, 50, or 75 cm wide on 1-m2 plots graded to < 0.5% slope were compared with that from a broadcast preemergence application. Herbicide concentration in runoff decreased with time after initiation of runoff and as application band width decreased. Band width reduced herbicide concentration in runoff to a greater extent than time after initiation of runoff. The 50-cm band treatment reduced atrazine and metribuzin loss 69 and 71%, respectively, from the broadcast treatment. Forty-nine percent less metolachlor was lost from the 50-cm band than from the broadcast treatment. Herbicide concentration or loss in surface runoff did not differ between the 50- and 25-cm band width treatments. Herbicide lost in runoff ranged from 4 to 7% of application with the smaller loss associated with narrower band widths. The greater reduction in atrazine and metribuzin loss in surface runoff than that associated with the reduction in input alone may be attributed to leaching or adsorption of herbicide by soil in the nonsprayed strip as runoff flowed from the treated to the nontreated areas. Band application of herbicides may provide an effective, consistent means to reduce herbicide loss in surface runoff.